Fracture networks in shale-hosted igneous intrusions: Processes, distribution and implications for igneous petroleum systems

Igneous intrusions in sedimentary basins can influence basin-scale fluid flow and petroleum systems in various ways. They may act as barriers, preferential pathways or even reservoirs for fluids. The fracture networks of intrusions usually represent the main control of their hydraulic properties. However, our understanding of different fracturing mechanisms and their quantitative effect on fracture network properties remains limited, and good field examples are sparse. Here, we present a comprehensive field study from the Neuqu ´ en Basin, Argentina, using a reservoir-scale outcrop of a sill complex emplaced in organic-rich shale, which constitutes a direct analogue of oil-producing fractured igneous reservoirs. We provide field evidence of various fracturing mechanisms affecting the fracture network, including cooling joints, bituminous dykes, hydrothermal veins, and tectonic faults. Using high-resolution digital fracture network quantification, we then tie these fracture mechanisms to spatial variations of fracture orientation, intensity and connectivity. Our results indicate that all observed fracture types are involved in hydrocarbon migration and/or storage. Bitumen of very high thermal grade within the intrusions implies migration of hydrocarbons into the sills in a destructive high-temperature environment. Importantly, bitumen dykes and faults locally alter the fracture network, creating zones of strongly increased fracture intensity and connectivity and therefore improved reservoir properties.


Introduction
Igneous intrusions are present in many sedimentary basins around the world.Their potential effects on geological processes and properties in the subsurface are widely recognized both on the local and regional scale.These effects include for instance accelerated chemical reactions (e.g.organic matter maturation) in the surrounding metamorphosed host rock and altered hydraulic properties, with consequences for groundwater exploration, CO2 sequestration and petroleum systems (Schutter, 2003;Senger et al., 2017), but also global climate change in the geological past (Iyer et al., 2017;Svensen et al., 2004).
With respect to fluid flow, igneous intrusions may act as barriers (Rateau et al., 2013), or alternatively, preferred pathways or even reservoirs (Mao et al., 2020;Rodriguez Monreal et al., 2009;Senger et al., 2015;Spacapan et al., 2020a), and may thus be of great importance for basin-scale flow patterns and hydrocarbon accumulations.Due to the typically low primary porosity of intrusions, fracturing represents a key factor controlling the overall hydraulic properties of igneous intrusions (Heap and Kennedy, 2016;Senger et al., 2017;Witte et al., 2012).Specifically for igneous intrusions acting as hydrocarbon reservoirs, fracture networks represent the main contributors to both porosity and permeability that allow storage and production in the first place (Delpino and Bermúdez, 2009;Senger et al., 2017).However, there is no consensus on the relative importance of different fracturing mechanisms for establishing pathways or storage capacity for fluids, or their effect on quantifiable fracture network properties.
Cooling joints constitute the most obvious element of the fracture network in igneous intrusions and are commonly believed to provide most of the fracture porosity and permeability in intrusions (Delpino and Bermúdez, 2009;Gudmundsson and Løtveit, 2014;Witte et al., 2012).Additionally, qualitative fracture evolution models often propose that fault-related fracturing and re-opening of cooling joints, as well as hydrothermal veins may enhance permeability (Spacapan et al., 2020a;Witte et al., 2012).Hydraulic fracturing due to fast maturation of organic-rich shales surrounding igneous intrusions is a well-recognized process (Aarnes et al., 2012;Zanella et al., 2015), but the potential fracturing of the intrusions themselves has been largely ignored.Although these existing models generally seem reasonable, there is an acute lack of published observational evidence of the specific fracturing mechanisms at play, and particularly their involvement in the hydrocarbon-bearing fracture porosity of igneous reservoirs.
Volcanological outcrop studies on cooling joints in intrusions propose a characteristic joint size depending on cooling rate, magma chemistry and intrusion geometry (Goehring and Morris, 2008;Hetényi et al., 2012).Others suggest power-law (i.e.scale-independent) behavior for several fracture parameters (length, aperture, orientation) for outcrop analogues of oil-bearing sills and assume laterally consistent fracture network properties (Witte et al., 2012).Connectivity is often believed to be high in cooling-joint networks, but this statement is frequently provided without quantitative evidence, particularly in the context of igneous reservoirs (Delpino and Bermúdez, 2009;Gudmundsson and Løtveit, 2014;Witte et al., 2012).Significantly, the existing models fail to address the strongly varying effect of igneous intrusions on fluid flow and storage.This has been reported both on the global scale (barrier vs. pathway, Senger et al., 2017) and locally, e.g. for intrusions acting as hydrocarbon reservoirs in the northern Neuquén Basin, Argentina, which exhibit strong lateral variations in their hydraulic properties (Spacapan et al., 2020a, A. Medialdea, pers. comm.).Thus, additional work is required to quantify fracture network properties in igneous intrusions and document lateral variations.Specifically, there is a pressing need to link this quantification to the underlying fracturing processes in order to improve our understanding of the fracture network evolution and potential causes of the observed hydraulic property variations of igneous intrusions.
Here, we present a comprehensive field study from the Neuquén Basin, Argentina, using a reservoir-scale outcrop of a sill complex emplaced in organic-rich shale, which constitutes a direct analogue of oil-producing fractured igneous sills.We show explicit field evidence of the wide spectrum of fracture types and mechanisms affecting the fracture network of an igneous intrusion emplaced in organic rich shale, and combine this with high-resolution digital fracture network quantification over more than 200 m.Our aims are (1) to provide sound evidence for the broad range of fracture types contributing to the accessible fracture network of igneous sills potentially acting as reservoirs, and (2) to quantify local variability in the fracture network parameters arising from the presence, or absence, of these fracture mechanisms.Thereby, we aim at providing new insight into fracture network evolution within igneous intrusions, with a specific focus on settings where the host rock is organic-rich shale.

Geological setting
The study area is located in the Sierra Azul range in the northern Neuquén Basin, Argentina, around 70 km south of the town Malargüe (Fig. 1a).The Neuquén Basin represents one of the Andean foreland basins and comprises a prolific hydrocarbon province with various proven hydrocarbon plays, including commercial oil production from fractured andesitic sills emplaced in organic-rich shales (Howell et al., 2005;Spacapan et al., 2020a).
The geodynamic evolution of the basin comprises three main phases (Fig. 1b).Initiation of the basin started in the Triassic-Jurassic as a series of elongated rifts forming isolated depocentres (Howell et al., 2005;Yagupsky et al., 2008).During the early Jurassic period, it subsequently transformed into a back-arc basin dominated by regional thermal subsidence following the onset of the Andean subduction.This phase lead to formation of the marine sediments of the Mendoza group, which include organic-rich shales of the Vaca Muerta and Agrio formations (Bettini and Vasquez, 1979;Manceda and Figueroa, 1995) that constitute two of the main source rocks of the basin.In the late Cretaceous, the tectonic regime changed to compressive, initiating the foreland basin phase.This stage created a series of fold-and thrust belts including the Malargüe fold-thrust belt through reactivation and inversion of the Rift-related normal faults as well as older basement faults (Manceda and Figueroa, 1995).The outcropping El Manzano sill complex described on this study was brought to surface during this phase and is located adjacent to one of these inverted halfgrabens, the El Manzano-Liu Cullín lineament (Yagupsky et al., 2008).
During the foreland basin stage, particularly the northern part of the Neuquén Basin experienced two main pulses (late Oligocene to Miocene; late Miocene to Pliocene) of extensive volcanism and magmatic intrusion of the sedimentary succession (Combina and Nullo, 2011;Kay et al., 2006).This magmatism profoundly affected the petroleum systems in several ways.First, it provided the necessary heat to generate hydrocarbons from the Vaca Muerta and Agrio formations, which are mainly immature in the study area as shown by thermal modeling and geochemical analysis (Spacapan et al., 2018).Secondly, it generated intrusions that themselves act as naturally fractured reservoirs, constituting atypical petroleum systems such as the Rio Grande Valley (RGV) oil fields (Witte et al., 2012;Spacapan et al., 2020a).The El Manzano sill complex presented in this study represents a direct outcrop analogue of O. Rabbel et al. the RGV petroleum system, as it is located just 10 km west of the oil fields, separated by a major thrust fault (Fig. 1a).The El Manzano outcrop features an exposed seismic-scale sill complex emplaced within the Agrio and Vaca Muerta formations, with interconnected andesitic intrusions of 1-20 m thickness (Fig. 2).Ar-Ar dating yields an age of 14 Ma for one of the sills at El Manzano (see supplementary data).This outcrop is spectacular not only due to its function as a direct analogue outcrop, but because it's continuous high-quality exposure for several kilometers allows studies ranging from the scale of seismic surveys (e.g., Rabbel et al., 2018) to detailed outcrop investigations.Some 5 km from the El Manzano sill complex, Borello (1944) first described outcrops of solid bitumen.Dykes and sills of bitumen (also termed "asphaltene") are common in the northern Neuquén Basin and have been mined extensively (e.g., Cobbold et al., 1999).Their formation is linked to fracturing and short-distance migration in response to fluid overpressure (Parnell and Carey 1995).The causes of overpressure formation have been debated for decades and are still uncertain, but strong heating pulses during phases of volcanism have repeatedly been proposed, because bitumen occurrence clusters around volcanic centers and the heat may trigger fast maturation and overpressure build-up (Rassmuss, 1923;Cobbolt et al., 2014;Zanella et al., 2015).

Field documentation, samples and geochemical analyses
The observations and data presented in this study where obtained during two campaigns totaling six weeks of fieldwork at the El Manzano outcrop, including exploration as well as detailed sampling (Fig. 2).In addition to the overall architecture of the sill complex, we focused on documenting the wide range of fracture types observed throughout the locality.This included drone surveys (DJI Phantom 3), field photographs, manual structural measurements of fault, fracture and bedding planes.We also collected 27 samples of different types of veins for hand sample and thin section descriptions, as well as Raman spectroscopy and qualitative conductivity measurements using a handheld multimeter setup to constrain the thermal alteration of organic matter in terms of potential graphitization (sample locations indicated in Fig. 2).Since solid, low-maturity bitumen is an insulator but graphite is an excellent conductor, any measurable conductivity in dry samples should indicate significant graphitization.Raman spectroscopy includes analysis of the positions and intensities of the spectral peaks at 1345 cm − 1 and 1585 cm − 1 , termed D ("disorder") and G ("graphite") peaks, respectively (Potgieter-Vermaak et al., 2011).For samples undergoing initial thermal alteration (prior to long-term thermal metamorphism), the G peak narrows and shifts towards 1615 cm − 1 with increasing alteration but shifts back to ca. 1598 cm − 1 if graphite is formed (Muirhead et al., 2012).Experimental investigations of short, intense heating of meteorite and immature bitumen samples indicate a rise in the peak intensity ratio I D /I G from around 0.6 to around 1, which may be used to investigate settings involving local heating due to igneous intrusions (Muirhead et al., 2012;Zhou et al., 2014).
Along a sub-vertical section through the main outcrop cliff (white lines in Fig. 2), we analyzed 16 shale samples using Rock-Eval pyrolysis.This is a standard tool to evaluate the realized and overall hydrocarbon generation potential of source rocks, and thus routinely used in the petroleum industry (e.g., Lafargue et al., 1998).RockEval serves as an indicator of thermal maturity and type of organic matter and is commonly used to quantify the maturation effect of igneous intrusions on organic-rich shale (Aarnes et al., 2011;Einsele et al., 1980;Spacapan et al., 2018).During pyrolysis, samples are heated to several hundred degrees in a step-wise process to identify the amount of free hydrocarbons (S1 peak), the remaining hydrocarbon potential (S2 peak), oxygen index (OI), total organic carbon (TOC) and hydrocarbon index (HI) (Espitalié et al., 1985).HI is calculated as S2/TOC x 100, yielding milligrams of hydrocarbons per gram of organic matter.Assuming initial HI from immature samples or literature values, this is used to calculate the transformation ratio (TR) of organic matter into hydrocarbons.

Fracture network quantification
This study aims to quantify potential lateral changes in fracture network parameters in response to the evidence of different fracture processes observed in the field.In addition to manual 1D scanlines for ground-truthing (Guerriero et al., 2010), we collected drone-based and handheld imagery to obtain high-resolution 3D digital outcrop models and orthophotographs from a selected outcrop location (Fig. 2) using structure-from-motion (e.g., Westoby et al., 2012).Augmenting the O. Rabbel et al. previously created large-scale, decimeter-resolution 3D model to interpret large-scale sill geometries (presented in Rabbel et al., 2018), the new 3D model allows mapping of fractures down to the sub-cm scale over a 230 × 20 m outcrop section.The model was georeferenced relative to the existing differential GPS-referenced large-scale model, leading to an integrated digital outcrop dataset across the scales.
In addition to collecting manual data, we extracted 3D plane orientation data of fracture planes along digital scanlines using the Lime software package (Buckley et al., 2019).Scanlines run both parallel and orthogonal to the intrusion boundary in order to minimize scanline orientation bias.This also exemplifies the benefit of digital outcrops in fracture mapping, as manual collection of fracture plane data from the inaccessible cliff in the intrusion center was not feasible.Note that for digital plane measurements, we used models with significantly higher mesh resolution compared to standard visualization purposes to avoid erroneous measurements resulting from smoothing of the digital outcrop surface.All fracture plane orientations were corrected for local dip of the surrounding sedimentary layers.
As a key element of the fracture network characterization, we generated an ultra-high resolution (down to 3 mm/pixel) orthorectified photograph of a 230 m outcrop section with exceptional outcrop quality and with many observations evidencing the presence of different fracturing processes (Fig. 2).We used the orthophotograph to manually pick 2D fracture trace maps, which subsequently served as input for automated fracture network analysis using the FracPaQ Matlab toolbox (Healy et al., 2017).This software provides comprehensive measurements and estimates of fracture network properties, including fracture density and intensity, connectivity, trace and segment length distributions, as well as 2D orientation distribution of the fracture network.2D fracture density and intensity, commonly termed P20 and P21 (e.g., Zeeb et al., 2013), represent number of fractures and length of fractures per unit area, respectively, and are measured here using the circular estimator method (Healy et al., 2017).Similar to the P10 value for 1D fracture intensity, commonly obtained from scanlines, P21, in particular, serves as an estimate for the fracture area per unit volume (P32), which is a critical input in fracture network modeling controlling the number of fractures in a model.
Connectivity of fracture networks is quantified using the network topology in terms of relative abundance of node types, i.e. crossing (Xnode), abutting (Y-node) and isolated termination (I-node).Note that any 2D mapping of fractures aimed at quantifying connectivity therefore needs to take care to correctly map fracture connections corresponding to the observed node types.For a given number N of X-,Y-and I-nodes, we get the average number of connections per line as C L = 4 * (NX+NY) NI+NY (Sanderson and Nixon, 2018).We calculated C L along a moving window of 35 m width that was shifted by 15 m, yielding a continuous curve of 15 connectivity values along the outcrop, each allocated to the position of the center of the window.
Finally, the FracPaQ provides a Maximum-Likelihood-Estimator (MLE) test to assess the likelihood of observed trace lengths being drawn from a (i) power-law, (ii) log-normal or (iii) exponential distribution (Rizzo et al., 2017).Thereby we aim to (1) avoid misleading results for length distribution characteristics from a wrong initial assumption, and (2) gain insight into how different fracture processes may alter the fracture length distribution.We performed this analysis in three manually selected subsets of the fracture network, according to outcrop-based observations of distinct processes, i.e., we separated the central area affected by a fault and bitumen dykes from the outer areas dominated by cooling joints.In order to account for potential truncation bias when fitting a power-law, we applied a lower cutoff according to the recommendations of Bonnet et al. ( 2001), while we did not apply an upper cutoff, because the intrusion gives a natural upper length limit through its size.
As a trade-off between data size and quality, we integrated two different drone surveys conducted from slightly varying distance to outcrop, resulting in different resolutions of 3 and 12 mm/pixel, respectively.This can be problematic, as the recorded number of fractures is often subject to resolution bias (Zeeb et al., 2013).However, we justify our approach by comparing digital results to manual scanlines (ground-truthing) and considering on-site outcrop observations (cf.supplement S1).Comparison of manual fracture density measurements with values from digital data in areas of lower resolution found the values to be identical, i.e. we do not expect significant resolution bias at the lower end of the scale.These areas show no signs of tectonic deformation or natural hydraulic fracturing, with small-scale fracturing absent upon inspection in the field.Where these processes are in fact present, we have high-resolution data available and are thus able to map smaller fractures, but without compromising comparability of the overall fracture data.

Cooling joints
As can be expected for thin sheet intrusions within a fold-thrust belt, we observe that the intrusions display strong, though not homogenous, jointing.Note that without additional observations (e.g., polygonal arrangement or orientation relative to intrusion-host contact), it is inherently difficult to distinguish cooling joints from tectonic joints.Therefore, some of the observed jointing may be of tectonic nature or at least may have formed in a stage of active compressive tectonics.However, since some additional observations are available and cooling joints represent a ubiquitous feature of igneous sheet intrusions, we refer to the observed joints as cooling joints.
Fig. 3 shows examples of cooling joints with bitumen filling or impregnation of their fracture surfaces.This is frequently found within intrusions at El Manzano.We occasionally observe polygonal joint arrangement typical of cooling joints, especially in places where the roof of an intrusion is exposed (Fig. 3a).In addition, the polygonal fracture planes are filled with a mixture of black and white material, identified as solid organic matter and calcite, respectively (Fig. 3a, close-up).In some cases, the joints exhibit plumose structures (Fig. 3b).The solid bitumen filling shows no evidence of leftover volatiles and instead appears as either a shiny black powder or bitumen (Fig. 3b and c).

Bituminous injection structures
Throughout the El Manzano sill complex, we observe large structures of injected bitumen or bituminous material within and around intrusions (Figs. 4 and 5; see Fig. 2 for sampled localities).Note that throughout this work, we use the term "bitumen" if veins or joints are filled or impregnated with pure bitumen, while "bituminous" describes observations where such fillings are impure (e.g.mixed with shale) or we are unsure of the exact nature of the organic matter.These injection structures occur in the form (1) arrays of pure bitumen dykes with observed individual sizes of up to ca 20 m height and 50 cm width (Fig. 4a and b), or (2) areas of intense and chaotic fracturing, where fractures are filled with a mixture of bitumen, shaly material and/or calcite (Fig. 5).These spectacular features are localized around the sills and originate in the aureoles from where they may extend far into the igneous bodies (Fig. 4a).We do not observe such dykes outside of the direct vicinity of an intrusion or its aureole, i.e. they do not seem to cut the stratigraphy over more than 10-20 m.On closer inspection, the dykes consist not of a single mass, but several branches, which may also cut across each other (Fig. 4b and c).In other cases, the dykes can mingle with thick calcite veins (see section 4.2.3),typically without a clear crosscutting relationship (Fig. 4d and e).While we did not find an exposed root of a pure bitumen dyke, the shale-bitumen filled fractures displayed in Fig. 5 appear to originate directly from the contact zone, where we observe meter-scale cavities due to missing host rock (Fig. 5a  and b).
O. Rabbel et al.The zones around such injection structures are intensely fractured and host large amounts of mobilized organic matter.While the fractured intrusive rock around the injection structure hosting bitumen-shale mixture disintegrates so readily that sampling is impossible (Fig. 5c), we were able to collect several samples of the more solid bitumen dykes (e.g.Fig. 4d, f, g).The hand sample shown in Fig. 4d displays mingling of cm-scale calcite and bitumen veins within intrusion.In addition, we observe that the bituminous material in many of the dykes has a shiny, fibrous, crystalline-looking appearance (Fig. 4f).In some of the samples, we found shale fragments and calcite crystals of millimeter to centimeter size in the bituminous matrix of such dykes (Fig. 4g).Note that due to the crosscutting of different types of veins, it is not always possible to keep the description of these features fully separated.
Fig. 6 depicts Raman spectra of three bitumen samples collected from intermingling bitumen and calcite veins in the southern part of El Manzano.All spectra show clear disorder and graphite peaks, of which the graphite peaks at around 1585 cm − 1 are particularly well developed.Peak intensity ratios I D /I G lie between 0.85 and 0.93.In addition, qualitative conductivity measurements indicate that dry bitumen samples with fibrous and shiny appearance conduct electrical currents.
RockEval pyrolysis provides results for thermal maturity of the Agrio formation at the El Manzano outcrop (Fig. 7).Assuming initial HI of 538 mg/g corresponding to the observed maximum value, our analysis indicates 0-20% transformation of organic matter in the parts of the section that are least affected igneous sills, while the aureole of the sills clearly shows increased transformation of up to 99% close to the sills, where injection structures are found.Note that other authors report lower initial HI in the area (e.g., Spacapan et al., 2018), which would lead to even smaller TR, especially in the less affected regions.

Hydrothermal calcite veins
Throughout the outcrop, we find many large carbonate veins cutting through both aureole and sills, that are distinctly different from calcitefilled cooling joints, and also show no signs of shear displacement (Fig. 8).The veins occur both at sill tips, and along the more central part of sills and may crosscut, or be crosscut by, bitumen dykes (Fig. 8a and b, also Fig. 5e).Commonly, these features are several centimeters to a meter wide, tens to hundreds of meters long, and include arrays of branching veins (Fig. 8a, c), and carry fragments of host rock (Fig. 8d).Millimeter-to centimeter-scale macropores like the one displayed in Fig. 8e are very common and often filled with bitumen, and release a strong hydrocarbon smell when broken up.Fluid inclusions with CO2 were identified using Raman spectroscopy (supplement S2).Occasionally, we find stains of highly viscous oil within the veins (Fig. 8f).

Faults and associated damage zones
Although the El Manzano outcrop is located in the hanging wall of the major thrust front of Sierra Azul range, faulting within the outcropping sill complex is limited to small-scale strike-slip and thrust faults cross-cutting the sills, as summarized in Fig. 9. Displacement along these faults typically is on the centimeter to meter range and locally causes enhanced fracturing in the form of damage zones.These damage zones commonly feature fractures filled with solid bitumen and low-viscosity, strongly degraded oil as well as calcite veins (Fig. 9b-d).Fig. 9c shows an example of a bitumen-filled fracture cross-cutting an amygdale partly filling primary vesicular porosity.Fig. 9e presents an example of a small-scale thrust fault cutting a sill and the surrounding sediments, which also contain bedding-parallel calcite veins.Despite the small extent of this fault, its core hosts a damage-zone filled with

Orientation patterns of fracture types
As the first step of a quantitative description of the variability of fracture types and their distribution, we provide dip-corrected plane orientations of different fracture types collected throughout the El Manzano sill complex (Fig. 10).Cooling joint data show predominantly vertical to subvertical planes.Despite a generally large spread in azimuths, there exists some agreement with the orientation of the main tectonic structures of the Sierra Azul range in the form of one E-W striking set and two conjugate NE-SW/NW-SE striking sets.Fault plane orientations show a NNE-SSW striking reverse fault and two ESE-WNW striking strike-slip faults, of which one is slightly oblique and has a reverse-fault component (cf.Fig. 9a).Bitumen dykes are characterized by vertical to subvertical dip and a pronounced preferred orientation in N-S and E-W direction, and thus with the large-scale tectonic structures.Orientations of the predominantly vertical hydrothermal calcite veins comprise one strong E-W striking set, but the remaining plane measurements spread over a wide range of orientations.

Influence of sill architecture
Overall, we observe that fracture patterns change notably with sill architecture and geometry.Although a full documentation of the sill Fig. 6.Raman spectra of bitumen obtained from samples of intermingling calcite and bitumen veins within the sills in the southern part of the outcrop (cf.Fig. 2).complex architecture and geometries found at El Manzano and their influence on the fracture network is beyond the scope of this work, the example in Fig. 11 qualitatively illustrates some key effects.The digital outcrop model illustrated in Fig. 11 is located in the northern part of the outcrop and comprises amalgamated and stacked sills of varying thickness.The massive sill in the lower left of Fig. 11 is approximately 11 m thick and terminates in this location, where the sill tip is visible.It comprises few, long subvertical joints as well as radial joints, which are characteristic for sill tips (see e.g., Galland et al., 2019).On the other hand, the even thicker sill on the right side consists of four stacked intrusions, which are distinguished through different weathering colour, clear horizontal contacts, and occasionally even thin lenses of host rock between them.Here, we find more, but shorter fractures mostly confined to one of the thin intrusion layers.
The P21 fracture intensity distribution (Fig. 12d) varies considerably along the outcrop.The outer domains (0-90 m and 160-230 m, respectively) comprise P21 values of 0-2.5 m -1 , averaging 1.2 m -1 .We find higher values in the central domain between 90 and 160 m, which includes the fault as well as the bitumen dyke, injection structure and intrusion step.Here, average P21 is 2.0 m -1 , with minimum values of around 1 m − 1 and local maximum values of up to 3 and 6 m − 1 around the fault and injection structures, respectively.
Finally, Fig. 12e illustrates the connectivity variations along the outcrop section in terms of average connections per line (C L ).We include theoretical thresholds for onset of initial fracture connectivity (C L = 2, red line) as well as percolation (C L = 3.57, green dashed line) for reference.Again, we identify that connectivity in the two outer domains differs markedly from the central domain.In the outer domains, we obtain connectivity values of 1.47-2.47and 2.12-2.44,respectively, i.e. partly below the threshold for a connected network.In the central domain, values vary between 3.25 and 4.02 average connections per line, i.e. the network is well connected and partly percolating.

Trace length statistics and MLE results
Statistical analysis results of the fracture trace length are presented in Fig. 13.For a qualitative overview, we display the fracture trace map color-coded by length Fig. 13a.It is already possible to see that the left and right domains (0-90 m, 160-230 m) are dominated by fewer, longer fractures compared to the central domain comprising fault and injection structures in addition to cooling joints (90-160 m).Also, we generally observe more and shorter fractures close to the contacts compared to the center.The overall trace length histogram indicates that small fractures dominate the mapped population with a peak at approximately at 0.4 m, and more than 80% smaller than 3 m.Interestingly, there is a small second peak at around 17 m, leading to a strongly asymmetric bimodal distribution.
We aim to quantify the lateral variability of the length distribution and assess the best mathematical description.Therefore, we present results of MLE fitting for power-law (Fig. 13c-e) and log-normal distributions (Fig. 13f-h) organized in three columns corresponding to the left, center and right parts of the outcrop, respectively, with a lower cutoff applied to account for truncation effects (red dotted line, supplement S3 for details).Table 1 summarizes the estimated likelihood for both distributions as well as exponential distribution, both with and without a cutoff.As a reminder, the MLE method estimates the likelihood that a population of values is drawn from a specific distribution.
For a power-law as well as exponential distribution, irrespective of an applied correction for truncation effects, the result is zero or nearzero percent likelihood (Table 1).For the power-law analysis shown in Fig. 13c-e, this is reflected in a visibly poor fit in all three domains of the outcrop.We also observe a clear "kink" resulting from a change in slope for fractures large than around 15 m, especially in the outer domains where cooling joints dominate (Fig. 13 c, e).Note in general that only the central domain data still stretch significantly more than one magnitude if a possible truncation bias is accounted for.
The MLE results for log-normal distributions yield ≥99 percent likelihood for the left, central and right domains of the outcrop, respectively, if no cut-off is applied to the observations (Table 1).Although generally applying a cutoff decreases the MLE likelihood, the domains are affected differently.While outer domains still reach values of 75 and 95 percent, respectively, the central domain drops to less than 5 percent likelihood, essentially ruling out a log-normal distribution.The generally better fit of log-normal distributions to our trace length data is visible in Fig. 13 f-h.Note again that in the left and right outcrop domains, the largest fractures (ca.15-20) m produce a local deviation from the trend, leading to underestimation by log-normal distributions (Fig. 13f, h).Exponential distributions score near-zero percentages in all MLE analysis.

Fracture network evolution
Our field observations show that, in the same geological setting, a wide range of fracture types can occur in the fracture network of igneous intrusions.We find bitumen and/or dead oil in all four identified fracture types within the sills at El Manzano (Fig. 3a and b; Fig. 4; Fig. 5b  and c; Fig. 8e and f; Fig. 9b-d), demonstrating that all fracture types may be accessible as pathways or even storage for hydrocarbons, but also fluids in general.Although most fracture types are directly related to the intrusion event, the underlying physical mechanisms of their formation are fundamentally different, resulting in a complex evolution of a laterally heterogenous fracture network.
Except for tectonic fractures, the fracture system in the sills is established within the period of cooling of the intrusion.While cooling joints form due to thermal stresses as the intrusion solidifies and contracts, the case of injection structures and large calcite veins are less obvious.However, the generally high thermal grade of the bitumen dykes in a geological setting with low burial-related temperatures and nearly no maturation (Fig. 7), as well as entrainment of host rock pieces in intermingling bitumen and calcite veins (Figs.4e and 8d) strongly suggests that these features formed in the high-temperature environment with substantial fluid-overpressure surrounding the intrusions.
The driving processes for early fracture formation in the intrusions are thus not tectonic but related to thermally controlled processes around and within the intrusion, such as cooling-related contraction, hydrothermalism and metamorphic reactions including organic matter maturation in the surrounding host rock.However, these processes occur within tectonic far-field stresses, which is reflected in the alignment with the tectonic structures and inferred E-W maximum compressive stress for all fracture types (Fig. 10).We will provide a more detailed interpretation of the individual fracture types in separate paragraphs.
Tectonically driven fracturing represents the only fracture mechanisms independent of the intrusion-related thermally controlled processes.Although overprinting due to faults is limited in the case of El Manzano, both qualitative and quantitative results show that faults may locally alter the initial fracture network and affect multiple intrusions (Figs. 9 and 12).The exact timing of faulting after emplacement remains unclear in this case.It is only constrained by the time of sill solidification and the end of tectonic activity in the Sierra Azul range, which are both Miocene.However, due to the short time required for the relative thin intrusions to cool (tens to hundreds of years), we generally regard faulting as a post-emplacement process.

Cooling joints
Cooling joints are ubiquitous features throughout the field area and comprise bitumen filling in many locations (Fig. 3).Their distribution and characteristics within the intrusions of the sill complex are quite variable and depend on several factors, including intrusion thickness, architecture and geometry.Fig. 11 illustrates the relationship between these factors and the cooling joint distribution.Thick, massive intrusions emplaced as a single body tend to develop fewer, but longer cooling joints compared to the many small fractures seen intrusions that may be of similar overall thickness, but consist of multiple, partly disconnected branches.This is most likely the result of different cooling rates in the two scenarios, since higher cooling rates produce smaller joints (Hetényi  et al., 2012).Even within intrusions the spatial distribution may vary, as we often observe more small cooling joints along the intrusion contact compared to the interior (Fig. 12b).This could be related to a similar effect of faster cooling of the intrusion boundary in the initial phase after emplacement, when temperature gradients are still large.Since cooling joints propagate perpendicular to the intrusion surface, any curvature induces a change in local orientation distribution, e.g. at sill terminations or steps (Figs.11 and 12).All of the above leads us to the interpretation that the emplacement mechanism and its controlling factors play a major role in the final cooling joint distribution, as it controls the shape and architecture of intrusions and thus defines the boundary conditions for joint formation.In addition, far-field stresses may lead to preferential alignment of the cooling joints with the prevailing stress field (e.g., Maher 2020), as the confining stresses are likely anisotropic (Fig. 10).
Due to the impregnation of joint surfaces with organic matter interpreted as high-grade bitumen (Fig. 3a and b), we suggest that cooling joints may act as fluid pathways for hydrocarbons when the intrusion is still in the cooling phase.Any fluids flowing through newly formed cooling joints would be heated to temperatures of several hundred degrees, likely leading to combustion of all volatile hydrocarbons, leaving solid leftovers in the form of bitumen.

Bituminous injection structures
The bitumen dykes and injection structures (Figs. 4 and 5) are perhaps the most spectacular fracture type observed in this study.We interpret them to result from strong heating and subsequent fast hydrocarbon generation in the metamorphic aureole of the sill intrusions, causing hydrofracturing within the shale due to fluid overpressure, sometimes up to the point of disintegration of the sediments.We base this interpretation on a combination of several observations.The very low background maturity level (TR = 0-20%) of the Agrio formation shale only increases to high values towards the intrusion contacts, where essentially all organic matter has been transformed to hydrocarbons (Fig. 7).Consequently, mobilization of hydrocarbons in the amounts necessary to form up to meter-thick bitumen dykes (Fig. 3) requires the heat input provided by the intruding magma.In the high-temperature environment of thermal aureoles, maturation takes place over very short periods (hours to years, e. g.Iyer et al., 2017;Panahi et al., 2019), which can be regarded as geologically instantaneous.This would not allow for flow-driven overpressure relaxation within the practically impermeable shale formations, leading to propagation of fractures.Strong additional fracturing of the intrusions, thinning and branching fractures, as well as partially missing host rock next to injection structures near the contact evidence propagation of the hydrofractures from the aureole into the intrusions (Figs. 4,5,and 12a,b).Mutual crosscutting of bitumen dykes or calcite veins shows that new fractures were repeatedly generated with enough pressure to propagate through the previous generation in a "crack-seal" fashion (Fig. 4c, e).Although previously formed cooling joints may have been exploited in some places, the propagation of bitumen dykes through the host rock and into the sill (Fig. 4a and b and 5) must be related to hydraulic fracturing.
Both Raman spectrograms (Fig. 6) and electrical conductivity of the bituminous material in the injection structures evidence strong thermal alteration and graphitization.This indicates formation of the injection structures shortly after solidification of the sill intrusions because this was the only period when temperatures required to form graphite were present at El Manzano (potentially >400 • C for hydrothermal graphite, cf.Buseck and Beyssac, 2014).
Since the bitumen dykes commonly consist of several branches that may also crosscut each other, we suggest that the dykes formed in a series of pulses (Fig. 4a-e).In addition, mutual cross-cutting of calcite veins and bitumen as well as calcite crystals entrained in bitumen dykes (Fig. 4d-g, Fig. 8a) point to (1) a repeated, violent fracturing process in accordance with the hydrofracturing model and (2) involvement of both hydrocarbon and other hydrothermal fluids.We interpret the calcite crystals in Fig. 4g as entrained rather than grown in-situ because of their random orientations within the bitumen dyke (as opposed to e.g.fibrous growth) and the pieces of shale visible in the same dykes, which are clearly not in-situ.The two sets of orientations of bitumen dykes (N-S and E-W, see Fig. 10) may indicate that the hydrofracturing occurred within the far-field stresses related to E-W Andean compression that created the Sierra Azul range.The relatively wide spread of orientations within the sets could be due to exploitation of, or interaction with, existing cooling joints, which show a wide scatter of orientations (Fig. 10).

Hydrothermal calcite veins
Due to their frequent occurrence and size, the large calcite veins observed throughout the El Manzano outcrop represent another important type of fractures present in the igneous intrusions (Fig. 8).The calcite veins observed within and at the tip of intrusions show (1) no indication of shearing and (2) mutual cross-cutting with the graphitic bitumen dykes (Fig. 4d and e and 8a).They are therefore not faultrelated, and likely formed contemporaneously with the bitumen dykes in a hydrothermal environment.Samples from calcite veins show inclusions of CO2 within the calcite crystals, which could have been provided either by the intrusion itself, or by degradation of hydrocarbons (supplement S2).Overpressure-driven hydrofracturing during intrusion-related hydrothermalism may be a potential cause of vein formation, since pieces of host rock are commonly entrained in the calcite veins, pointing to a process violent enough to break up the host rock and transport the pieces over some distance (Fig. 8d).Additionally, the observation of bedding-parallel fibrous calcite veins may indicate overpressure in the metamorphic aureole (Fig. 9e).Since the calcite veins often crosscut intrusions (with single veins occasionally crosscutting several intrusions) and can reach many decimeters in thickness, we infer that they are the result of vertical movement of hydrothermal fluids at the scale of tens or even hundreds of meters.Quantitative fracture network characterization along a 230 m long outcrop section reveals strong parameter variations caused by local interaction of several fracture types and mechanisms (Figs. 12 and 13).The mapped outcrop shown in Fig. 12 has three domains which comprise different fracture types and features.While the outer domains (0-90 m, 160-230 m, Fig. 12a) are dominated by cooling joints, the central domain (90-160 m) includes a fault and two injection structures, as well as a primary step in the intrusion geometry.The step is identified by offset in the sill, but an absence of offset in the sedimentary layer above.While the properties are similar in the outer domains with cooling joints only, the addition of other fracture types in the central domain leads to an increase in density, a wider spread in orientations, and much higher connectivity (Fig. 12b-e).We suggest that the interaction of fracture types other than cooling joints does not only simply add more fractures, but leads to a wider orientation distribution, since faults and injection structures do not necessarily follow paths perpendicular to the intrusion contact like cooling joints.As increasing connectivity requires not just more fractures, but also fracture intersections, the addition of oblique fractures by injection structures and faults is vital for an actual connectivity increase.Importantly, the step in the intrusion geometry introduces inclined cooling joints, which further diversifies the fracture orientations.
Additionally, the lateral variability in the presence of different fracture types also manifests itself in the fracture length statistics (Fig. 13).The qualitative impression from the colored fracture map is that smaller fractures are concentrated in the central domain (Fig. 13a).Thus, instead of focusing on the overall length distribution for the entire outcrop (displayed as a histogram in Fig. 13b), we analyze the length distribution per domain.Here, it becomes clear that the outer domains with mainly cooling joints show very similar characteristics.The distributions comprise fractures between ca.0.2 and 20 m, with few small and large joints, that fit well with log-normal distributions according to MLE analysis (Fig. 13c, e, Table 1), but not power laws or exponential distributions (Fig. 13f, h, Table 1).We interpret this result in the light of the existence of a characteristic length scale for cooling joints that depends on cooling rates and the size of the intrusion, especially since other fracture types that may alter the distribution are absent.From this perspective the poor fit with a power law is unsurprising since powerlaws imply scale-independent distributions.It is noteworthy that the largest joints spanning the entire intrusion thickness (15-20 m) create a "kink" in the distributions that leads to underrepresentation in the lognormal distribution.This indicates that a log-normal distribution also does not fully capture the geological process.
On the other hand, the results from the central domain are more difficult to capture in terms of mathematical descriptions.Despite a striking increase in smaller fractures due to faulting and injection structures, MLE analysis yields a poor fit with a power-law even for biascorrected data, which is also obvious from a qualitative comparison of the best fit and the data (Fig. 13g, Table 1).However, MLE only favors a log-normal distribution if no lower cutoff is applied, and in fact discards a log-normal distribution otherwise (Fig. 13d, Table 1).Based on outcrop observations, we believe that truncation bias of some extent exists in this area, suggesting that not all fractures are represented even in the high-resolution models and a correction should be applied.Consequently, we interpret the lack of any reasonable fit with standard mathematical models common in fracture length description to be a result of interacting fracture processes with very different underlying physics.The blend of cooling joints, a fault-related damage zone and two injection structures creates a highly heterogenous local fracture population which markedly increases fracture density and connectivity but does not seem to follow any unified fracture length distribution.

Fracture network evolution
Establishing the fracture evolution within the igneous intrusions at El Manzano is an important step towards understanding fracturing of igneous intrusions in organic-rich shales in general.In addition, it provides further insight into the igneous petroleum system evolution of producing igneous sill reservoirs, such as those in the Neuquén Basin.Our multi-disciplinary methodology allows us to constrain timing and temperature conditions during fracture formation, and demonstrate the quantitative manifestation of the wide spectrum of fracture processes in the form of laterally varying fracture parameters over hundreds of meters.In the following paragraphs, we will utilize our results to critically evaluate the existing conceptual model of thermal impact and fracture evolution in the igneous petroleum systems in the Rio Grande Valley (Spacapan et al., 2020a;Witte et al., 2012).Based on this evaluation, we present a more process-oriented model focused on fracturing mechanisms that honors both previous knowledge and the plentiful field evidence provided in this article (Fig. 14).

Thermal phase: hydro-thermo-mechanical interaction
The existing models emphasize the importance of intrusion-related processes.Spacapan et al. (2020a) distinguish two phases: The initial "thermal stage" is dominated by temperature-driven metamorphic reactions in the aureole and fluid migration away from the intrusion due to local pressure gradients.During the subsequent "cooling stage", a cooling joint network is established and hydrothermally driven hydraulic fracturing occurs in the host rock, leading to a first pulse of fluid and hydrocarbon migration into the intrusions due to a reversed fluid pressure gradient (Witte et al., 2012).This model implies a clear temporal sequence of fluid generation in high temperature environment, followed by fracturing and migration in a cooler environment.
However, from our observations, a key characteristic of the early phase after magma emplacement is that processes appear to take place simultaneously, and chemical reactions, fracturing and migration have an immediate causal relationship.The occurrence of injection structures showing strong thermal alteration (graphitization) and evidence of forceful injection , as well as bitumen impregnating cooling joints (Fig. 3) and hydrothermal veins carrying host rock pieces and intermingling with bitumen dykes (Fig. 8) demonstrate that hydrocarbon-bearing fluids entered the intrusion at temperatures of several hundred degrees.The intense modification of the fracture network (Figs. 12 and 13) we observe within intrusions through interaction of bituminous injection structures and cooling joints shortly after emplacement represents a novel result that is not included in previous models.
Our interpretation of emplacement of injection structures in distinct pulses fits well with results from field studies, laboratory models and numerical studies of maturation-related fracture networks, where repeated pressure build-up and release is observed (Kobchenko et al., 2014;Panahi et al., 2019;Rabbel et al., 2020).Previous field studies reported on extensive bitumen dykes in the northern Neuquén basin and suggested a connection to volcanic activity based on general proximity of the dykes to volcanic bodies and radial bitumen dyke orientations around volcanic centers (Rassmuss, 1923;Cobbold et al., 2014;Zanella et al., 2015).Our study demonstrates this connection and, at least in the study area, clearly establishes that the bitumen dykes formed before the intrusion has fully cooled down.Note, however, that both the thermal maturity and proximity to volcanic structures of bitumen dykes across the basin varies greatly.Different modes of formation have been suggested and their evolutional model requires a case-by-case evaluation (Parnell and Carey 1995).
Our interpretation agrees with observations from other studies of sediment-intrusion interactions focusing on pressure evolution in these systems.From the moment of emplacement, the intrusion cools down and shrinks, progressively generating primary porosity in the form of cooling joints and vesicles as well as strong fluid underpressure even before complete solidification (Aarnes et al., 2008).At the same time, temperatures in the organic-rich shales rise.The resulting fast hydrocarbon generation, mineral dehydration and thermal fluid expansion causes a drastic rise in pore fluid pressure in the impermeable shale, leading to hydraulic fracturing (Aarnes et al., 2012).The occurrence of bedding-parallel fibrous veins (Fig. 9e) in the contact aureole as well as bitumen dykes at El Manzano (Figs. 4 and 5) matches well with previous work from the Neuquén Basin that related these features to strong heating and hydraulic fracturing in shales, possibly due to volcanic activity (Zanella et al., 2015).Thus, a situation arises in which different fracturing processes within sediments and intrusions are ongoing, and fluid flow of newly created fluids or fluidized sediments into the intrusion (c.f.Fig. 14 a-e) is facilitated by an inward-directed fluid pressure gradient (Svensen et al., 2010).Importantly, the intrusion temperature at this time is still at several hundred degrees, causing thermal alteration of the incoming material.
Overall, the dynamic period directly after emplacement appears to be defining for the formation of the initial fracture distribution in intrusions emplaced in organic-rich shales.Despite the comfort afforded by a model displaying the clear distinction of phases by timing and processes, we argue that a more accurate description of the "thermal phase" of an evolutionary model should comprise simultaneous, interacting processes leading to an initially heterogeneous fracture network, as depicted in Fig. 14.Tectonic overprinting after the intrusion has cooled down may further alter this fracture network, potentially also establishing connections to previously isolated primary pores (Figs.9c and 14).

Fracture network properties of igneous reservoirs
Our field observations indicating direct evidence for bitumen and/or dead oil occurring in each fracture type demonstrate that all types of fractures found in the igneous intrusions can be involved in hosting hydrocarbons (Fig. 14).This complements previous conceptual studies based on well data and modeling, which focus on the contribution of cooling joints to the fracture porosity of igneous reservoirs (e.g., Delpino and Bermúdez, 2009;Gudmundsson and Løtveit, 2014).We present an easily accessible, world-class locality where geologists can observe the role of different fracture types in the petroleum system firsthand.High lateral variability of fracture intensity and connectivity (Fig. 12) implies strong changes in fluid storage capacity as well as fracture permeability over short distances.In this context, interaction of multiple fracture types appears to enhance reservoir quality locally.Additionally, our results highlight laterally varying fracture length statistics and render power-law behavior for fracture length in igneous intrusions unlikely (Fig. 13, Table 1).Among the available choices evaluated using the MLE method, a log-normal distribution with laterally varying parameters provides the best results, although in intensely fractured areas affected by multiple fracturing mechanisms this model fails to accurately reflect the entire fracture network.This may be explained by an overprinting of fracture populations associated with the different fracturing processes, some of which (e.g., tectonic fracturing) may create a power law distribution, while others (e.g.cooling joints) have characteristic length scales related to intrusion geometry and boundary conditions of the cooling process (Goehring and Morris, 2008;Hetényi et al., 2012).Overall, the striking lateral variability of the fracture network over 10s-100s of meters contrasts with previous findings by Witte et al. (2012), who stated that fracture networks and therefore permeability should be relatively constant over such distances.We therefore advise that any attempt to model the fracture networks of intrusions in a petroleum system should include careful evaluation of evidence for lateral heterogeneity of the fracture network.
Production data from oil fields close to El Manzano (Fig. 1) were published by Spacapan et al. (2020a), showing high initial production rates followed by a sharp drop, sometimes only after a couple of months.They related this to initial drainage of the hydrocarbons in the fracture network, followed by a much smaller mixed matrix-fracture production.As the El Manzano sill complex represents a direct outcrop analogue for the adjacent subsurface systems of the Rio Grande Valley comprising comparable extent and intrusion sizes, the outcrops may serve to develop a better understanding of the subsurface fracture systems holding the producible hydrocarbons.Perhaps, the local areas of extreme fracture density seen at El Manzano (Fig. 12) represent the pockets of high fracture density that are drained in the initial phase of production.However, previous research pointed out the existence of so-called cavity zones that may not always be related to fracturing but could also be due to large vesicles form by late-stage volatiles (Bermúdez and Delpino, 2008;Witte et al., 2012).A promising next step would be to utilize the quantitative outcrop observations to conduct fluid flow and production simulations for comparison.

Hydrocarbon generation and migration
The thermal effect of intrusions in terms of a metamorphic aureole with rapid hydrocarbon maturation is well established through case studies around the world (Aarnes et al., 2010;Muirhead et al., 2017;Rodriguez Monreal et al., 2009;Spacapan et al., 2018).In the case of El Manzano, this effect is an overall positive one, since previously immature source rocks show elevated transformation ratios around the intrusions (Fig. 7).However, the discovery of graphitic bitumen dykes and bitumen within intrusions and their aureoles is an important discovery, because it provides evidence for mobilization and migration of hydrocarbon-bearing fluids into the intrusions while the temperatures are sufficient for substantial thermal alteration.Although our thermal methods are qualitative and cannot assess the exact temperature, the observed graphitization indicates that the hydrocarbon-rich fluids experienced "in-situ combustion" at several hundred degrees.Considering the impressive size of some the observed bitumen dykes, a significant part of the petroleum potential around the intrusions may be lost in the first pulse of local fluid generation and migration, which has previously been proposed as a charge mechanism for oil into the intrusions (Spacapan et al., 2020a;Witte et al., 2012).It is important to note that the bitumen dykes observed in the outcrops are always proximal to intrusions and do not extend far from the contacts, making them very localized migration features.Assuming a penny-shape for the bitumen dyke shown in Fig. 4a as lower boundary, this feature alone comprises around 160 m 3 (ca.1000 barrels) of solid bitumen.Further investigation should explore this migration pulse in more detail to quantify the hydrocarbon volume that may survive.It seems reasonable to assume that for liquid hydrocarbons to survive and be produced, migration into the fractured intrusions must take place after the intrusion has fully or nearly cooled down to ambient temperatures.
Furthermore, our observations agree with the suggestion of Spacapan et al. (2020a) that both the faults and hydrothermal calcite veins may connect different levels of sill intrusions and facilitate vertical migration of hydrocarbons from lower stratigraphic levels, in this case the Vaca Muerta Fm (Fig. 14).At El Manzano, both features commonly intersect and thus connect several intrusions, and show evidence of hydrocarbons occupying their porosity (Fig. 8e and f, 9c and d).These features may therefore control local fluid flow patterns and should be included in reservoir-scale flow modeling.

Implications for geophysical properties
Finally, our results may reveal some of the geological mechanisms responsible for the large variations in geophysical properties observed in well data from the intrusive reservoirs near the study area (Rabbel et al., 2018;Spacapan et al., 2020b;Witte et al., 2012).In addition to mineral composition (both primary and secondary), porosity and fracturing strongly control seismic velocity changes in igneous rocks (e.g., Berge et al., 1992;Mark et al., 2018).Although we do not provide measurements or modeling of potential fracture-related velocity changes in the intrusions investigated in this study, the drastic increase in fracture density as well as the associated alteration should be capable of reducing seismic velocities of intrusions significantly.In fact, the El Manzano field locality previously served as a site for outcrop-based seismic modeling of a sill complex investigating the effects of velocity variations on seismic imaging of intrusions and found marked dimming of reflections if velocities are at the lower end of the spectrum (Rabbel et al., 2018).In addition, increased fracturing due to metamorphic reactions in the aureole as well as the occurrence of graphite may contribute to the low-resistivity zones around sills documented by Spacapan et al. (2020b).Thus, the presented field study may give applied geophysicists a more practical understanding of the geological features they are dealing with and illustrate potential causes of the wide spectrum of seismic responses that are possible in igneous intrusions.

Conclusions
We present an interdisciplinary field study of the fracture network in an igneous sill complex emplaced in organic-rich shale in the northern Neuquén Basin, Argentina.Our observations including drone and ground based photogrammetry and geochemical data demonstrate that a wide spectrum of fracturing mechanisms can affect the fracture network of intrusions.We employ digital fracture mapping on highresolution digital outcrop models to quantify the effect of these processes on fracture network properties.Based on the results, we draw the following conclusions: • We identify four different fracture types present in sills, driven by a broad spectrum of physical processes.These include (1) cooling joints created by thermal contraction, (2) faults and associated tectonic fractures, (3) bitumen dykes and bituminous shale injection structures, and (4) hydrothermal veins, likely due to hydrofracturing initiated in the metamorphic aureole.• Except for tectonic fractures, the fracture network is established during the early phase dominated by a high-temperature environment.Evidence suggests that in this phase, hydrocarbons generated may be mobilized and migrate into the intrusion and experience temperatures of several hundred degrees, where in-situ combustion of volatiles and graphitization of solid bitumen takes place.• We observe the different fracture types throughout the outcrop, but they do not occur spatially homogenous.Instead, the fracture network properties exhibit strong lateral changes on the scale of some tens to hundreds of meters, depending on the number and intensity of processes involved.Such changes involve locally increased fracture density, wider orientation distribution, higher connectivity and a variable length distribution.• Comparison of different mathematical distributions shows that fracture lengths are best represented through a spatially varying lognormal distribution, although even this is not overall satisfactory.Our analysis also demonstrates that previously suggested power-law behavior is very unlikely and difficult to justify considering the various physical processes at work.• Strong variations in fracture network properties imply similar heterogeneities for petrophysical and geophysical properties, such as locally enhanced storage capacity and fracture permeability or reduced seismic velocities.
Overall, this study extends our understanding of fracture networks in igneous intrusions acting as petroleum reservoirs and provides numerous field examples and quantitative data.The El Manzano field analogue can thus be used help geoscientists of all disciplines to better constrain modeling efforts and develop a more practical understanding of subsurface igneous petroleum systems.

Fig. 1 .
Fig. 1.(a) Satellite image of the study area in the Sierra Azul range, adjacent to the producing fractured igneous reservoirs of Rio Grande Valley, with indicated Andean E-W compressive stress regime.(b) Regional tectono-stratigraphic framework (modified from Spacapan et al., 2020a).

Fig. 2 .
Fig. 2. 3D view of the El Manzano sill complex (intrusions indicated in red), along with a wide-angle drone photograph (insert), demonstrating the outcrop scale and relevant locations of data presented in this study.Apparent clustering of features and absence in parts of the outcrop results from targeted field campaigns.(For interpretation of the references to color in this figure legend, the reader is referred to the Web version of this article.)

Fig. 3 .
Fig. 3. Examples of cooling joints within intrusions.(a) Intrusion roof with polygonal joint arrangement (see red weathering along surfaces), with fractures showing alteration and black and white fill (calcite and bitumen, see close-up).(b) Joint surface with plumose structure impregnated with shiny bituminous material, (c) irregular joint surface with bitumen.(For interpretation of the references to color in this figure legend, the reader is referred to the Web version of this article.)

Fig. 4 .
Fig. 4. Photographs of bitumen dykes from field and hand samples.(a) Large bitumen dyke crosscutting the intrusion-host rock contact and branching within the intrusion.(b) Bitumen dyke array within a metamorphic aureole.(c) Internal structure of the dyke shown in (a) with multiple fibrous dykelets.(d, e) Intermingling calcite and bitumen veins in a hand sample and at outcrop scale, respectively.(f) Fibrous, shiny sample of a bitumen dyke.(g) Calcite crystals and shale pieces entrained in the matrix of a bitumen dyke sample.

Fig. 5 .
Fig. 5. Bituminous shale injection structure.(a) Drone-based virtual model showing the sill geometry and spatial relation of the fractures and holes in the host rock.(b,c) Field photographs illustrating the missing host rock as the apparent origin of a large fracture filled with black material, surrounded by an intensely fractured zone.

Fig. 7 .
Fig. 7. Hydrogen index (HI), transformation ratio (TR) and TOC profile from Pyrolysis transect through the intruded Agrio formation.Green area represents lower and upper TR estimates based on reference values from Spacapan et al. (2018) and local maximum as the initial HI, respectively.(For interpretation of the references to color in this figure legend, the reader is referred to the Web version of this article.)

Fig. 8 .
Fig. 8. Large calcite veins in the field.(a) Intermingling calcite veins and bitumen dykelets at an exposed intrusion tip.(b) Drone based photograph showing a ca. 10 cm thick calcite vein cutting through aureole and sill.(c, d) Large, 1 m thick vein cutting the contact aureole of a sill with entrained host rock pieces.(e) Sample of a vein with bitumen occupying a cm-scale pore.(f) Calcite vein containing oil stains.

Fig. 9 .
Fig. 9. Examples of faults and associated damage zones observed at El Manzano.(a) Oblique strike-slip fault evident in a drone-based virtual outcrop model.(b) Field photograph of the damage zone associated with the fault displayed in image (a).(c, d) Hand samples showing bitumen and oil-filled fractures in the same damage zone.(e) Small thrust in a sill comprising a damage zone with thick calcite veins.Note also bedding-parallel calcite veins in the shale between the two intrusions.

Fig. 10 .
Fig. 10.Dip-corrected orientation data of fracture and fault planes sorted by fracture type (Schmidt equal-area projection).

Fig. 11 .
Fig. 11.(a) Digital outcrop model showing several amalgamated and stacked intrusive bodies.(b) Interpretation of the different intrusive bodies and layers based on the model and outcrop observations, with brown colours representing igneous bodies, and the grey background representing sedimentary host rock.(For interpretation of the references to color in this figure legend, the reader is referred to the Web version of this article.)

Fig. 12 .
Fig. 12. Fracture network quantification using a high-resolution digital outcrop model.(a) Orthophotograph indicating main geological features and scanline locations.(b) 2D fracture map of the outcrop.(c) Stereograms with poles of fracture planes measured along virtual scanlines on the 3D model.(d) Map of fracture intensity (P21).(e) Connectivity along outcrop based on fracture network topology.(For interpretation of the references to color in this figure legend, the reader is referred to the Web version of this article.)

Fig. 13 .
Fig. 13.Fracture trace length statistics.(a) Fracture map colored by trace length.(b) Overall trace length histogram.(c-e) probability density functions plotted in double-logarithmic, along with MLE best fit for a power-law distribution (black stippled line).(f-h) Probability density functions plotted in semi-logarithmic, along with MLE best fit lines for log-normal distribution.Red stippled lines indicate data range if cut-off is applied to account for potential truncation bias.(For interpretation of the references to color in this figure legend, the reader is referred to the Web version of this article.)

Fig. 14 .
Fig. 14.Schematic summary of fracture types, underlying mechanisms and fracture network variations interpreted from field evidence presented in this paper.Central image visualizes the large-scale features, including intrusion architecture and its relation to spatial fracture network variations.Insets (a-f) highlight specific fracture types and potential implications for igneous reservoirs.

Table 1
MLE percentage likelihoods sorted by outcrop spatial domain and tested distribution type.